How to extend an AI video past 30 seconds

To extend an AI video on Sume, generate a new clip whose first_frame is a still from the end of the previous one, then join the clips with Timeline 1.0. Sume's POST /v1/videos has no extend parameter; the vendors' own extension modes are documented separately.

BytePlus's Seedance 2.5 tutorial describes an extend task in ModelArk, and Google's Gemini Omni page describes an extension that appends up to a 40-second total.

How do the routes compare?

The ceilings differ by route.

How do I grab the last frame?

Import the clip into Sume, then call POST /v1/video-frames with a time near the end. It is unbilled, the source must be a media.sume.com clip, and at accepts up to 24 times. Use a time strictly less than the clip's duration.

curl -X POST https://api.sume.com/v1/video-frames \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: last-frame-001" \
  -d '{
    "video_url": "https://media.sume.com/artifacts/artf_demo/clip1.mp4",
    "at": [9.5]
  }'

How do I continue and join?

Send the returned frame as first_frame in frame_images on the next seedance-2.5 request, with a prompt that continues the action. Repeat, then render the sequence with Timeline; see assemble long-form video with the Timeline API.

Will the seam show?

A hard cut can show a change in lighting or motion. Match the ending pose in the last-frame still and the opening description of the next prompt, and use a Timeline transition such as dissolve.